8
$\begingroup$

I tried experimenting with some ab initio NMR calculations using the NWChem program. I used quinuclidine as a simple test case. The input file was:

echo
start quinuclidine_b3ylp_opt_nmr_shift

geometry
  c         -6.06576        0.97412       -0.15301
  c         -5.70979        1.98552       -1.28158
  c         -4.36490        2.65888       -0.91745
  h         -6.47615        2.78876       -1.35616
  h         -5.68491        1.45114       -2.25601
  c         -3.37563        1.54341       -0.50314
  c         -3.86410        0.94013        0.84625
  h         -2.33065        1.90335       -0.38494
  h         -3.37977        0.77674       -1.30967
  n         -5.29417        1.27496        1.08291
  h         -3.26800        1.37762        1.67961
  h         -3.67643       -0.15484        0.87417
  h         -5.79345       -0.05430       -0.48418
  h         -7.16227        0.96043        0.02635
  c         -5.40220        2.73070        1.37047
  c         -4.60680        3.55788        0.31857
  h         -5.06760        2.98401        2.39936
  h         -6.47931        3.01065        1.31755
  h         -3.96835        3.24945       -1.77113
  h         -3.61077        3.85719        0.71410
  h         -5.16983        4.48719        0.08363
end

basis
 * library 6-311G
end

dft
 xc b3lyp
end

title "Quinuclidine B3LYP/6-311G geom opt nmr shift"
task dft optimize

basis
 * library 6-311G
end
dft
 xc b3lyp
end

property
 gshift
 shielding
end

task dft property

Next, I calculated the chemical shift of the TMS standard:

echo
start tms_b3lyp_opt_freq_nmr_shift

geometry units angstroms
  si  -5.37335102864362      2.43869139121700     -0.00017138545369
  c   -3.50689688966580      2.43868715601638     -0.00001133876668
  c   -5.99529466705065      4.19430001974021      0.12209335653192
  h   -5.64408778702798      4.67985919142380      1.04819971762123
  h   -7.09781476514675      4.22936546250064      0.12467467866764
  h   -5.64435465416627      4.80382888763710     -0.72767673129762
  h   -3.10677010278341      2.89237170000804      0.92240456642084
  h   -3.10633327621776      3.01074016686707     -0.85387463583008
  h   -3.10622471891726      1.41321522592654     -0.06819713587793
  c   -5.99568159818493      1.45555000740651      1.45925444723856
  c   -5.99571944212990      1.66683194553091     -1.58150765777166
  h   -5.64537419258706      1.88774171172852      2.41177722402156
  h   -5.64408400205021      0.41082962309567      1.41778721376507
  h   -7.09821437068132      1.43547142531899      1.48770835119311
  h   -5.64584598058862      2.22683552653798     -2.46512226305553
  h   -7.09824537965055      1.65038972286304     -1.61218398701376
  h   -5.64368714450789      0.62658083618162     -1.68513442039297
end

basis
 * library 6-311G
end

dft
 xc b3lyp
end

title "TMS VWN5/aug-cc-pVDZ geom opt nmr shift"
task dft optimize

basis
 * library 6-311G
end
dft
 xc b3lyp
end

property
 gshift
 shielding
end

task dft property

However, the resulting chemical shifts are terribly off from the experimental ones. Am I doing an obvious mistake or is the big deviation just the best you can get out of such calculations? What is your experience and approach to ab initio NMR calculations (i.e. chemical shifts and coupling constants)?

$\endgroup$
  • $\begingroup$ I also tried out calculation chemical shifts/coupling constants with ORCA. However, in all the simple test cases I tried out, I always got huge deviations from experiment. This is why I am unsure now if I make an obvious mistake, e.g. not using a solvent model or if those calculations just are that unprecise. I somehow feel that just not handling solvent or using an appropriately sized basis set should completely screw up the results of such calculations. $\endgroup$ – logical x 2 Nov 17 '16 at 21:12
  • 1
    $\begingroup$ I know of some cases (such as Pt NMR) where the absolute isotropic shielding constant is EXTREMELY sensitive to the geometry (a deviation of 0.01 Angstroms can lead to a change in this value by a couple hundred ppm or more). However, you can sometimes get good error cancellation with your reference compound. Also, try considering solvent in your calculations (COSMO, PCM, etc.) though your mileage may vary. $\endgroup$ – LordStryker Nov 17 '16 at 21:57
  • 1
    $\begingroup$ It is hard to discern what exactly went wrong without a peak at the output files. Did you scale the values output? See this site for more information: cheshirenmr.info/index.htm $\endgroup$ – BiggChemT Nov 18 '16 at 0:25
  • 1
    $\begingroup$ Can you upload (or pastebin.com ) the relevant section of the outputs. I am curious about how they look like. I've never used NWChem nor Orca for chemical shift calculations. Thanks in advance. Also, notice that it can be important in order to answer your question. $\endgroup$ – user1420303 Nov 18 '16 at 0:48
  • $\begingroup$ @user1420303 I will do that ASAP. $\endgroup$ – logical x 2 Nov 18 '16 at 8:09
2
$\begingroup$

Those calculations are tricky because some functionals work well, others don't, some basis sets work well while even a better basis might give bad results and so on.

We did some studies in the past using the GIAO Method in G09 with B3LYP and 6-311+G(2d,p) als basis for the NMR calculations and M06-2X/6-311++G(d,p)for geometry optimization. And that worked quite well. However, we always used solvated systems (PCM) which seems to be the common way to do it. Try to reoptimize in solvent and run the NMR calculation in solvent if that's possible in NWChem.

$\endgroup$
  • $\begingroup$ I will perform the calculations with a solvent model and see if it gets better. Thanks : ) $\endgroup$ – logical x 2 Nov 18 '16 at 8:07
  • $\begingroup$ While I may lack specific experience in the calculation of chemical shifts, it is immediately apparent that "general purpose" and frankly outdated basis sets of the Pople variety such as 6-311G... are not a good choice. Things may improve if the basis set is (partially) decontracted (to give sufficient variational freedom at the nuclei) - but then one gives up any advantage of the small basis set and may use a better one anyway. $\endgroup$ – TAR86 Apr 10 '17 at 5:30
0
$\begingroup$

I use Gaussian 09 to obtain NMR $\ce{^{13}C}$ chemical shifts (cs) with great accuracy. The best results are obtain if the calculated cs are scaled. So, I recommend to use this level of theory:

  1. Opt => mPW1PW91/6-31G(d) with tight criteria (opt=tight)
  2. NNR => mPW1PW91/6-31G(d) with nmr int=ultafine
  3. To scale: $\text{scaled cs} = \text{(calculated cs)}*1.05 - 1.22$ (See reference)
$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.